Bioefficacy.AI: A Cutting-Edge Tool to Fast-Track Herbicide Formulation Development Using Photosynthesis-Based Bioefficacy Scores

Herbicides are one of the most vital tools in modern agriculture, playing a key role in protecting crop productivity by effectively managing weeds that compete for nutrients, water, and sunlight. Their use ensures healthier crops and maximizes yields, supporting food security and sustainable farming practices. 

However, the path to developing these vital crop protection tools is fraught with significant challenges. Creating new herbicides entails a complex and demanding process, involving the identification of effective active ingredients and the assurance of effective delivery. This requires comprehensive research, meticulous testing, and innovative formulation strategies. A pivotal aspect of this process is the selection of co-formulants, which not only ensures the stability and safety of the formulation but also enhances the delivery of the active ingredient to target weeds, maximizing efficacy and improving weed control. 

Results

After the formulation development phase, potential candidates must be rigorously evaluated for their effectiveness. One widely recognized method for this evaluation is bioefficacy measurement. Although this technique is relatively straightforward, as it involves visual analysis, it carries the risk of low accuracy due to its reliance on human interpretation, which can introduce subjectivity and variability into the results.

In addition to the risks of low accuracy, visual bioefficacy assessments are both time-consuming and considerably expensive. The process requires multiple replications to ensure reliable results, with each test taking between fifteen to thirty days to yield outcomes. This extended timeframe poses challenges for timely decision-making regarding formulation candidates, delaying progress and increasing costs.

To enhance collaboration with customers and partners in herbicide formulation development, Indovinya, the chemical specialties division of Indorama Ventures, has developed a fast-tracking tool to screen formulation candidates using artificial intelligence to predict bioefficacy scores. This innovative method is four times faster, consumes fewer resources, and drastically reduces workload compared to traditional screening methods, streamlining the development process and driving greater efficiency.

Bioefficacy.AI is based on deep photosynthesis understanding and the acknowledge of it as a methodology capable of predicting plant behavior days after the application of a formulated herbicide.

Why photosynthesis? 

For plant physiologists, it is the cornerstone of plant survival—a vital metabolic process that governs growth and adapts to stress conditions by signaling tolerance and resilience mechanisms. Under extreme stress conditions, such as those expected because of an herbicide application, photosynthesis is expected to exhibit stress responses in the form of senescence and death signaling.

Given its significance, measuring photosynthesis serves as an efficient method for screening herbicide formulations. Herbicides exert substantial stress on plants, triggering senescence and death signals in the photosynthetic process. While photosynthesis measurements do not pinpoint the mode of action, they effectively reflect the efficiency of weed control, akin to bioefficacy scores.

Photosynthesis measurements are conducted using an infrared gas analyzer (IRGA). This device features a chamber with precisely controlled concentrations of CO2 and water. By comparing the differences in CO2 and water concentrations between the chamber and the leaf, the IRGA provides a precise evaluation of the plant's photosynthetic activity, enabling accurate and reliable analysis. 

The IRGA system calculates over 20 variables with high precision and minimal variability. However, the complexity of interpreting these results necessitates multivariate analysis to discern patterns that reveal the most effective surfactant combination for a given herbicide. To streamline decision-making and transform photosynthesis data into a user-friendly format, we have incorporated artificial neural networks, enabling faster and more accurate insights.

Artificial Neural Networks (ANNs, Figure 1) are learning and solving complex problems techniques. ANNs are quick and simple in parameter estimation, highly adaptable to different situations, and have great potential to present more precise and coherent results compared to mathematical models or visual assessments. Therefore, the motivation is to obtain an Artificial Neural Network to optimize the data collection time for bioefficacy results based on initial photosynthesis data collected days after herbicide application in a targeted weed. 

Figure 1: Artificial Neural Network used in our technology.

Evaluating the Effectiveness of Bioefficacy.AI: A Case Study on Glufosinate Formulation Development

Eight formulations of ammonium-glufosinate (AG; SL formulation-type; 200 g/L; 3 g AI/ha) were applied to control Caruru-gigante (Amaranthus retroflexus L.) plants. Photosynthetic activity data was collected using an IRGA (LI-6800; Li-Cor Biosciences) within a 24 to 72-hour interval post-application. This data was then processed using our technology to assess formulation performance and efficacy.

Within just 72 hours of photosynthesis measurements, we can accurately predict bioefficacy scores for periods ranging from 7 to 28 days. This rapid analysis also allows us to utilize photosynthetic data to determine the optimal subdose for various formulations, significantly enhancing decision-making efficiency in herbicide development.

Figure 2: Predicted bioefficacy and real bioefficacy scores for 7 and 30 days after formulation application. * Significant differences between the real and the predicted bioefficacy. Letters indicate statistical differences between treatments (ANOVA, p<0.05; Tukey, p<0.05).

The predicted scores closely align with the actual bioefficacy results (Figure 3). Additionally, formulations 5 and 7 demonstrated the highest predicted bioefficacy, indicating stronger plant death signals and greater overall effectiveness.

Figure 3: Predicted bioefficacy and real bioefficacy scores for 14th days after formulation application. * Significant differences between the real and the predicted bioefficacy. Letters indicate statistical differences between treatments (ANOVA, p<0.05; Tukey, p=0.016).

As evidenced by the successful Glufosinate formulation screening case, Bioefficac.AI combines Indovinya’s expertise in advanced crop solutions with digital capabilities to transform screening herbicide formulation processes. Leveraging from our broad portfolio tool box and innovative products, this AI-powered tool predicts bioefficacy with precision, enabling customers to save time and reduce costs in the early stages of herbicide development. Technology we seed, PREDICT and harvest together.